Open Source Advocate Argues DeepSeek is 'a Movement... It's Linux All Over Again' (infoworld.com) 33
Matt Asay answered questions from Slashdot readers in 2010 (as the then-COO of Canonical). He currently runs developer relations at MongoDB (after holding similar positions at AWS and Adobe).
This week he contributed an opinion piece to InfoWorld arguing that DeepSeek "may have originated in China, but it stopped being Chinese the minute it was released on Hugging Face with an accompanying paper detailing its development." Soon after, a range of developers, including the Beijing Academy of Artificial Intelligence (BAAI), scrambled to replicate DeepSeek's success but this time as open source software. BAAI, for its part, launched OpenSeek, an ambitious effort to take DeepSeek's open-weight models and create a project that surpasses DeepSeek while uniting "the global open source communities to drive collaborative innovation in algorithms, data, and systems."
If that sounds cool to you, it didn't to the U.S. government, which promptly put BAAI on its "baddie" list. Someone needs to remind U.S. (and global) policymakers that no single country, company, or government can contain community-driven open source... DeepSeek didn't just have a moment. It's now very much a movement, one that will frustrate all efforts to contain it. DeepSeek, and the open source AI ecosystem surrounding it, has rapidly evolved from a brief snapshot of technological brilliance into something much bigger — and much harder to stop. Tens of thousands of developers, from seasoned researchers to passionate hobbyists, are now working on enhancing, tuning, and extending these open source models in ways no centralized entity could manage alone.
For example, it's perhaps not surprising that Hugging Face is actively attempting to reverse engineer and publicly disseminate DeepSeek's R1 model. Hugging Face, while important, is just one company, just one platform. But Hugging Face has attracted hundreds of thousands of developers who actively contribute to, adapt, and build on open source models, driving AI innovation at a speed and scale unmatched even by the most agile corporate labs.
Hugging Face by itself could be stopped. But the communities it enables and accelerates cannot. Through the influence of Hugging Face and many others, variants of DeepSeek models are already finding their way into a wide range of applications. Companies like Perplexity are embedding these powerful open source models into consumer-facing services, proving their real-world utility. This democratization of technology ensures that cutting-edge AI capabilities are no longer locked behind the walls of large corporations or elite government labs but are instead openly accessible, adaptable, and improvable by a global community.
"It's Linux all over again..." Asay writes at one point. "What started as the passion project of a lone developer quickly blossomed into an essential, foundational technology embraced by enterprises worldwide," winning out "precisely because it captivated developers who embraced its promise and contributed toward its potential."
We are witnessing a similar phenomenon with DeepSeek and the broader open source AI ecosystem, but this time it's happening much, much faster...
Organizations that cling to proprietary approaches (looking at you, OpenAI!) or attempt to exert control through restrictive policies (you again, OpenAI!) are not just swimming upstream — they're attempting to dam an ocean. (Yes, OpenAI has now started to talk up open source, but it's a long way from releasing a DeepSeek/OpenSeek equivalent on GitHub.)
This week he contributed an opinion piece to InfoWorld arguing that DeepSeek "may have originated in China, but it stopped being Chinese the minute it was released on Hugging Face with an accompanying paper detailing its development." Soon after, a range of developers, including the Beijing Academy of Artificial Intelligence (BAAI), scrambled to replicate DeepSeek's success but this time as open source software. BAAI, for its part, launched OpenSeek, an ambitious effort to take DeepSeek's open-weight models and create a project that surpasses DeepSeek while uniting "the global open source communities to drive collaborative innovation in algorithms, data, and systems."
If that sounds cool to you, it didn't to the U.S. government, which promptly put BAAI on its "baddie" list. Someone needs to remind U.S. (and global) policymakers that no single country, company, or government can contain community-driven open source... DeepSeek didn't just have a moment. It's now very much a movement, one that will frustrate all efforts to contain it. DeepSeek, and the open source AI ecosystem surrounding it, has rapidly evolved from a brief snapshot of technological brilliance into something much bigger — and much harder to stop. Tens of thousands of developers, from seasoned researchers to passionate hobbyists, are now working on enhancing, tuning, and extending these open source models in ways no centralized entity could manage alone.
For example, it's perhaps not surprising that Hugging Face is actively attempting to reverse engineer and publicly disseminate DeepSeek's R1 model. Hugging Face, while important, is just one company, just one platform. But Hugging Face has attracted hundreds of thousands of developers who actively contribute to, adapt, and build on open source models, driving AI innovation at a speed and scale unmatched even by the most agile corporate labs.
Hugging Face by itself could be stopped. But the communities it enables and accelerates cannot. Through the influence of Hugging Face and many others, variants of DeepSeek models are already finding their way into a wide range of applications. Companies like Perplexity are embedding these powerful open source models into consumer-facing services, proving their real-world utility. This democratization of technology ensures that cutting-edge AI capabilities are no longer locked behind the walls of large corporations or elite government labs but are instead openly accessible, adaptable, and improvable by a global community.
"It's Linux all over again..." Asay writes at one point. "What started as the passion project of a lone developer quickly blossomed into an essential, foundational technology embraced by enterprises worldwide," winning out "precisely because it captivated developers who embraced its promise and contributed toward its potential."
We are witnessing a similar phenomenon with DeepSeek and the broader open source AI ecosystem, but this time it's happening much, much faster...
Organizations that cling to proprietary approaches (looking at you, OpenAI!) or attempt to exert control through restrictive policies (you again, OpenAI!) are not just swimming upstream — they're attempting to dam an ocean. (Yes, OpenAI has now started to talk up open source, but it's a long way from releasing a DeepSeek/OpenSeek equivalent on GitHub.)
Seen repeatedly - collaboration / platforms - win. (Score:2, Insightful)
This is the thing we keep seeing. It's the reason the GPL won out over BSD. Everyone in BSD was always thinking "I'll suck in these people, get their help fixing my bugs then, just at the last minute I'll sucker them and sell it as proprietary". Because nobody will admit it but (almost) everyone's thinking the same thing (except a few like de Raadt - who honestly believes the stuff many BSD people just spout, which is why they all hate him) they are all watching and waiting to try to destroy the upstream pr
Re: (Score:1)
AleRunner (accidentally pressed AC it seems).
Re:Seen repeatedly - collaboration / platforms - w (Score:5, Insightful)
At that point everyone has state-of-the-art and a good position for developing their own package. Why would anyone buy software and support when the CEO/senior dev. just 'fired' all his staff?
So everyone except the USA can use this currently superior data-set. Worse, everyone else can contribute to making it better (and cheaper) than whatever data-sets the USA is selling. That's a brilliant plan from the USA government.
Re:Seen repeatedly - collaboration / platforms - w (Score:5, Interesting)
... BAAI on its "baddie" list.
This is why:
This democratization of technology ensures that cutting-edge AI capabilities are no longer locked behind the walls of large corporations or elite government labs but are instead openly accessible, adaptable, and improvable by a global community.
It's a direct threat
This is like Britain's Salt Act of 1882 which prohibited Indians from collecting or selling salt
Re: (Score:3)
Or Bectel's water subsidiary declaring to the citizens of Cochabamba, Bolivia, that they were no longer permitted to capture rainwater for irrigation or consumption in 2000. (Company executives had to flee the city for their lives.)
Re:Seen repeatedly - collaboration / platforms - w (Score:4, Insightful)
The purpose of the GPL is to use the world wide copyright protections as an enforcement mechanism for a more permissive license which enables unrestricted derivative works by all users.
I suppose, if you can find a country where copyright is not enforced, you could dispense with the provisions of the GPL, but it would still be needed in countries that enforce copyright laws.
The MIT license is not strong enough to prevent the closing off of derivatives.
Sooo... this is the year of ... (Score:5, Funny)
Re: (Score:2)
Your desktop can probably run deepseek-distill-llama-8b.
Re: (Score:2)
And here we thought the prophecies of old, that Linux would one day conquer the desktop, meant some Linux distro would conquer the desktop. Maybe the real meaning of the prophecies, was never about Linux distros, but will finally be fulfilled by DeepSeek!
Like it! (Score:5, Insightful)
Politicians (Score:4, Informative)
>"If that sounds cool to you, it didn't to the U.S. government, which promptly put BAAI on its "baddie" list"
It is pretty apparent over the years that politicians do not understand open source (nor the ramifications of many technologies).
Now, just because something is open source, doesn't mean it won't ever contain "bad" things- it entirely depends on who is participating in the project. But, unlike closed source, anyone can examine it and see what is in it and what it does, and can even fork it into a totally new project if needed. But that also doesn't mean people are actually monitoring/auditing the project. So it isn't clean and easy. A home-grown closed source project by a (as they put it) a non-"baddie" might also have compromised or bad parts in it, but you are not able to look.
Being open source doesn't guarantee it is "good", but if you care, you can find out, with ongoing effort. Being closed source doesn't guarantee it isn't "bad", but you can never really know- you can only look at inputs and outputs and have blind "faith" in the company/organization behind it.
What was he going on about? (Score:5, Insightful)
"it stopped being Chinese the minute it was released on Hugging Face"
No, it would stop being Chinese if other people generated a model without the same pro-CCP orientation baked in. It might resemble open source in some way if doing that was practical, even if no one has done it yet. But putting weights, a model card and even a puff-piece paper on Hugging Face do not come close to that.
"What started as the passion project of a lone developer [....]"
Is he confusing the Linux kernel for all of open source? I thought we got over that blinkered view of open source and free software 25 years ago.
Let's see some models that can actually be reproduced before claiming that LLMs are like open source software.
Re: (Score:3)
>"No, it would stop being Chinese if other people generated a model without the same pro-CCP orientation baked in. It might resemble open source in some way if doing that was practical"
That is a very good point. There is the project to use the model, and the model, itself. Although the project to use the model can be completely open, how is that even possible with a model- unless everything that was used to create the model were also open? And even if it were, which "normies" would have the resources
Re:What was he going on about? (Score:5, Insightful)
The open-source version isn't very biased. Use a system prompt that clearly states what you want (e.g., uncensored, unbiased, direct, not using too many emojis, etc.) and it works quite well. Don't expect their web service to be uncensored.
If it were, it wouldn't be available for long. In China, they have to play by the rules.
Re: (Score:1)
Very true. And, in China, it's always been true that he who has the gold makes the rules.
Re: (Score:1)
Re:What was he going on about? (Score:4, Informative)
They released everything, including the tools to train your own version.
The main limiting factor is that even with their super efficient method, it still costs a few million Euro to do.
But if you just want to be rid of the pro-CCP bias, you can simply do some brain surgery on their model and re-educate it.
Re: (Score:3)
> But if you just want to be rid of the pro-CCP bias, you can simply do some brain surgery on their model and re-educate it.
And there are simple methods to do so. People do it all the time so the American models don't moralize when they mention sex.
Re: (Score:2)
Is he confusing the Linux kernel for all of open source? I thought we got over that blinkered view of open source and free software 25 years ago.
No, he is not, why would you think that? He is talking about Linux kernel, how it blossomed as open source project - using it as an example of a successful open source project.
Re: (Score:2)
No, he is not, why would you think that?
I would think that because there are no good parallels between AI and the Linux kernel. AI models are not a single person's passion project, and they're not clearly on track to become core infrastructure. The leading models are closed source and too huge for hobbyists to run, or even most teams to run on-premises. The models that people can run for themselves are still more or less toys in comparison, and have stayed that way. That makes it much more difficult for a loosely coupled team to collaborate.
Re: (Score:2)
No, he is not, why would you think that?
I would think that because there are no good parallels between AI and the Linux kernel. AI models are not a single person's passion project, and they're not clearly on track to become core infrastructure. The leading models are closed source and too huge for hobbyists to run, or even most teams to run on-premises. The models that people can run for themselves are still more or less toys in comparison, and have stayed that way. That makes it much more difficult for a loosely coupled team to collaborate.
I would argue that even though Linux originated as a single person's passion project, it quickly became a joint effort, with ever growing contributor base.
Similar argument can be said on the SCIENCE behind neural networks. I read a good book on the history of the subject - since about 2010, as the interest started growing in neural networks, the leading minds insisted that their employers let them publish their findings (i.e. science). This is the reason that others have been able to build on that experti
US crypto export regulations (Score:2)
It really sounds more like the early days of US crypto export regulations than linux/OSS.
Trump is going to build a wall around America (Score:2)
because then we will be supergreat! Go us!
Linux? Hahahaha, no. (Score:2)
Linux is a kernel reimplementation of SYS-V and BSD with a GNU userspace. There was never any doubt that it would be quite useful and work well. LLMs are experimental stuff that is still in seatch of an application and they definitely do not work well at this time and nobody knows whether that will change.
Re: (Score:3)
Re: (Score:2)
The original versions of Linux compared to AIX or VAX was wheelbarrow to 18-wheel truck, too. Thirty years later it's the truck. I doubt it will take AI models that long.
Re: (Score:2)
Wrong. There was instances that _worked_ (AIX, Solaris, etc.) and Linux reimplemented them and copied their kernel API. There are no such instances for AI models.
Re: (Score:2)
but if they simply said that then they would not have made this crazy comparison that ended up on slashdot
And that is probably why they made that insane statement. Agreed.
attempting to reverse engineer (Score:5, Insightful)
"Hugging Face is actively attempting to reverse engineer and publicly disseminate DeepSeek's R1 model" sounds more dramatic than it is. They are using the public methods to build a new model.
DeepSeek did not only publish their model under the MIT license, but also described in the paper how to train it. Additionally, sometime after, they had an open source week where they released a lot of their tools as open source. DeepSeek is really not trying to hide what they did, but they are actively encouraging the open source community to build upon their foundation.
Don't Soil Linux with AI (Score:2)